Computers and Geotechnics 134 (2021) 104107

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

On the application of the beam model for linear dynamic analysis of pile
and suction caisson foundations for offshore wind turbines
Guillermo M. Álamo *, Jacob D.R. Bordón , Juan J. Aznárez
Instituto Universitario de Sistemas Inteligentes y Aplicaciones Numéricas en Ingeniería (SIANI), Universidad de Las Palmas de Gran Canaria, Edificio Central del Parque
Científico y Tecnológico, Campus Universitario de Tafira, 35017 Las Palmas de Gran Canaria, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: Piles and suction caissons are the most common foundation solutions for fixed Offshore Wind Turbines at in­
Offshore wind turbine termediate water depths. They are generally used as a single element, presenting large diameters and short aspect
Suction caisson foundation ratios. These specific dimensions drastically differ from the ones of classical applications (offshore platforms,
Monopile foundation
bridges, tall buildings etc.). Thus, in this paper the validity of their modelling as beam elements for the particular
Seismic response
problem of OWT is revised. The results of a soil-beam model, based on the integral Reciprocity Theorem in
Boundary elements
Finite elements Elastodynamics and specific Green’s functions for the layered half-space for the soil behaviour coupled with
Timoshenko’s beam Finite Elements, are benchmarked against the ones of a soil-shell model, based on Boundary
Elements for the soil coupled with shell Finite Elements. The comparative study is conducted in terms of
foundation characterization variables (impedance functions and kinematic interaction factors). Their influence
on the OWT seismic response is also studied through a substructuring procedure. From the results, some ex­
pressions for determining the applicability range of the beam simplification are proposed as functions of the
relative foundation-soil stiffness ratio. It is observed that this applicability range goes beyond that the one
commonly considered. .

1. Introduction
Pile and suction caisson foundations (also known as buckets, suction
piles or suction anchors depending on the context) are being used as
foundations of fixed Offshore Wind Turbines (OWT). Such solutions are
being considered at sites with shallow (10 to 30 meters) and interme­
diate water depths (30 to 60 meters) with single (monopile or mono­
bucket) or multiple foundation arrangements depending on both water
depth and soil properties.
As wind turbines becomes larger, the required piles and suction
caissons becomes bigger in diameter: up to 8 meters for monopiles (XXL
monopiles), up to 30 meters for caissons (Cotter, 2009), and even bigger
for shallow composite caissons with internal skirts (see e.g. Jia et al.,
2018). Such diameters are much bigger than diameters used in other
more classical projects (gas/oil platforms, bridges, buildings, etcetera),
and therefore the use of models traditionally considered should be taken
with care. The length L to diameter D ratios ranges are 3 < L/D < 10 for
monopiles and 1 < L/D < 6 for suction caissons (Houlsby and Byrne,
2005a; Houlsby and Byrne, 2005b). The buried part of both piles and
caissons are constituted by steel shells of thickness around t/D ∼ 0.01
(pile shaft) for monopiles and t/D ∼ 0.001 (caisson skirt) for suction
caissons.
The complete design of an OWT is a complex task which should fulfil
many requirements (DNV, 2014). Although rough designs can be ob­
tained from simplified procedures at early stages, see e.g. (Arany et al.,
2017), final designs should be defined after using an integrated opti­
mization procedure which guarantees a safe and economical solution,
see e.g. (Ashuri et al., 2014). In this sense, it is of fundamental impor­
tance an appropriate modelling which should also be computationally
efficient due to the many evaluations needed during the design process.
In the case of foundation modelling, recent works (Bhattacharya et al.,
2013; Bordón et al., 2019; Page et al., 2019), have highlighted its
importance particularly for the prediction of first and second natural
frequencies, which are relevant for performing the crucial fatigue ana­
lyses. In the case of seismic analyses, the frequency range usually goes
far beyond the second natural frequency. The use of convenient
simplifying hypotheses allows adopting simpler and cheaper models
(ideally via closed-form formulae or calibrated fast models) at initial
steps, whereas more rigorous, complex and detailed models should be
considered at the final stages.

* Corresponding author.
E-mail addresses: guillermo.alamo@ulpgc.es (G.M. Álamo), jacobdavid.rodriguezbordon@ulpgc.es (J.D.R. Bordón), juanjose.aznarez@ulpgc.es (J.J. Aznárez).

https://doi.org/10.1016/j.compgeo.2021.104107
Received 7 October 2020; Received in revised form 25 January 2021; Accepted 25 February 2021
Available online 21 March 2021
0266-352X/© 2021 Elsevier Ltd. All rights reserved.
G.M. Álamo et al.
Fig. 1. Problem layout and substructuring model.
The present paper is concerned with the modelling of big diameter
and short piles and suction caissons for seismic analyses of OWT sup­
ported by a single foundation element (monopile or monobucket). The
focus is put on assessing the adequacy of different simplifying hypothesis
which allow the use of more elementary models from which results are
more easily obtained. To this end, a previously developed soil-beam
coupled model (Álamo et al., 2016) is compared against a rigorous
soil-shell model (Bordón et al., 2017). The former is a model which re­
duces the soil-structure interaction to a line (pile axis), and greatly re­
duces the number of degrees of freedom. The pile itself is modelled with
Timoshenko finite elements, and the soil response is included via
Green’s function for an arbitrary horizontally layered half-space, which
avoid the discretisation of the free-surface and the layer interfaces, i.e.
only pile axis nodes are present. The latter model reduces the soil-
structure interaction to the shell mid-surface, which leads to a
rigorous and general interaction model but it is also computationally
costlier. The comparative study is performed in terms of impedances and
kinematic interaction factors. In addition to this, results in terms of OWT
variables are also compared in order to evaluate to what extent the
differences in the foundation modelling are transmitted to the whole
tower-supporting structure-foundation system.
The use of some kind of one-dimensional reduction, i.e. beam-like
models, for foundation modelling is quite advantageous for obvious
mathematical and computational reasons. The key of such models is how
soil-structure interaction is considered. Pioneering works take the
rigorous Mindlin’s solution and integrate it along the pile axis or pile
shaft under certain assumptions (see e.g. Jiménez-Salas and Belzunce,
1965; Thurman and D’Appolonia, 1965; Poulos and Davis, 1968). An
alternative approach is the well-known Winkler model, where distrib­
uted springs and dashpots are connected along the pile axis (Novak
et al., 1978; Gazetas and Dobry, 1984). Advanced approaches use some
Green’s function under point or ring loading to introduce the soil
response, and then this is coupled to beam finite elements in different
ways (Kaynia and Kausel, 1982; Coda et al., 1999; Almeida and de Paiva,
2004; Padrón et al., 2007). The soil-beam model used in this paper ex­
tends (Padrón et al., 2007) by considering an arbitrary horizontally
layered half-space and Timoshenko finite elements. The aim of the
present work is to study to what extent this soil-beam model, initially
developed for pile foundations, is able to reproduce soil-structure
2
Computers and Geotechnics 134 (2021) 104107
interaction (impedances and kinematic interaction factors) for length
to diameter ratios as low as 1, i.e. suction caissons.
The paper is organized as follows. First, Section 2 presents the two
different foundation models together with the substructuring model
used to compute the OWT system response. Then, the comparison be­
tween the beam and shell models is conducted along the different parts
of Section 3. The comparison starts in terms of the foundation response:
static (Section 3.1) and and dynamic (Section 3.2) stiffness, and kine­
matic interaction factors (Section 3.3). Then, some expressions deter­
mining the upper frequencies below which the beam model accurately
reproduces the foundation response are proposed in Section 3.4. The
final part of the comparison, in terms of OWT variables, is done through
the application example presented in Section 3.5. The paper ends listing
the main conclusions drawn from the study in Section 4.
2. Methodology
2.1. Substructuring model
The problem at hand is summarized in Fig. 1: an Offshore Wind
Turbine connected to a submerged structure founded on a large diam­
eter and relatively short monopile or suction caisson subjected to
vertically-propagating S waves.
A two-dimensional (lateral behaviour) substructuring model is used
for modelling the OWT dynamics. It comprises a simple concentrated
mass at the hub representing the Rotor-Nacelle-Assembly (RNA), a suf­
ficient number of Euler–Bernoulli beam elements for taking into account
the conical tower and the submerged part. The foundation dynamics is
synthesized via frequency dependent springs and dashpots (impedance
functions) connected to the submerged part base. The foundation energy
filtering when impinged by a seismic action (vertically incident shear
waves) is taken into account via kinematic interaction factors. An
adequate representation of both impedances and kinematic interaction
factors is vital for the study of OWT modal and seismic behaviour.
The impedance functions relate the forces and moments and the
displacements and rotations of the foundation at the mudline level. They
are frequency-dependent complex functions whose real and imaginary
parts represent respectively the stiffness and damping characteristics of
the foundation. Given that the study focuses on the lateral behaviour,
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107

Fig. 2. Foundation configuration and its modelling: (a) steel hollow pile in homogeneous half-space, (b) DBEM-FEM model (Bordón et al., 2017) (soil-shell inter­
action), (c) Integral model (Álamo et al., 2016) (soil-beam interaction).

only horizontal KH , rocking KR and sway-rocking cross-coupling KHR

impedance functions are considered in the substructuring formulation. 2.2. Soil-shell model
However, the vertical stiffness term KV is also analysed in this work
when studying the foundation-only response. The monopile shaft and the suction caisson skirt are topologically
The kinematic interaction factors represent the filtering effects of the similar, and the only differences are the dimensions of the foundation
foundation, and they are computed as the ratio between the displace­ element. In both cases the thickness to diameter ratios t/D are well
ment or rotation at the foundation and the free-field motion at the below 5%, being smaller in the case of suction caissons skirts (typically
mudline. The translational and rotational kinematic interaction factors t/D ∼ 0.001) than in the case of monopile shafts (typically t/D ∼ 0.01).
are respectively denoted as Iu and Iθ . The consideration of a purely continuum three-dimensional solid model
More details about the methodologies (soil-shell or soil-beam) used for the soil-foundation system is therefore unnatural and overly com­
to evaluate the impedance functions and kinematic interaction factors plex. Instead, the monopile shaft or suction caisson skirt is more
are given in next sections. In both models, a time harmonic analysis at appropriately modeled as a shell, leading to a mixed dimensional model
circular frequency ω is considered, and the soil is assumed to be a ho­ which considerably reduces the required number of degrees of freedom.
mogeneous linear elastic solid with following properties: shear modulus Such a model has been developed by the Authors (Bordón et al.,
Gs , Poisson’s ratio νs , density ρs and hysteretic damping ratio ξs ; being 2017) via a BEM-FEM approach which considers the interaction be­
the complex effective shear modulus to be used G*s = Gs (1 + i2ξs ). tween the foundation and the soil as the interaction of a shell (FEM) and
The governing equations in the frequency-domain that are used to the surrounding soil (BEM), see Fig. 2. The main hypothesis of this
obtain the system response can be written as: model is thus the reduction of soil-shell interaction to the mid-surface of
([ ss ] [ ss ] ){ s } { } the shell. The present model is fully described for the case of Biot’s
K Ksb M Msb u 0 poroelastic soils in (Bordón et al., 2017), but a brief description of the
− ω2 = (1)
K bs bb
K +K f
M bs
M bb
ub Fb
this model for an elastic soil is outlined below.
The soil domain Ωs is discretized by using the BEM, which is based on
where u is the vector containing nodal in-plane lateral displacements
the use of Boundary Integral Equations (BIE) relating displacements uk
and rotations, K and M are the stiffness and mass matrices of the system
and tractions tk throughout its boundary Γ = ∂Ωs . The boundary Γ
obtained by assembling the elementary ones, the indexes distinguish
consists of two parts: free-surface boundary Γfs , and shell mid-surface
between all structural nodes (s) and the base (mudline level) one (b), Kf
Γshm considered as a crack-like boundary (Γshm = Γ+ shm + Γshm ). The
−
is the impedance matrix of the foundation, and Fb is the force and Singular BIE is used for collocating along the free-surface Γfs (Domí­
moment acting at the base of the structure due to the seismic excitation. nguez, 1993):
This force vector can be computed from the impedance matrix and ∫ ∫
kinematic interaction factors as: cilk uik + t*lk uk dΓ = u*lk tk dΓ (4)
{ } Γ Γ
[ ] Iu
Fb = K f (2)
Iθ where l, k = 1, 2, 3 and the Einstein summation convention is implied.
For comparison purposes, the response of the system neglecting Soil- The tensor cilk is the free-term at the collocation point, uik is the
Structure Interaction (SSI) effects is also considered, i.e. rigid base displacement at the collocation point, and u*lk and t*lk are the elastody­
assumption. In this case, the free field motion uff is directly introduced at namic fundamental solutions in terms of displacements and tractions
the mudline node of the superstructure, while restricting its rotation. respectively. The Dual (Singular and Hypersingular) BIEs are used for
This simplification reduces Eq. (1) into: collocating on the shell mid-surface Γshm :
{ } ∫ ∫
( ss ) ( ) 1 ( i+ )
K − ω2 Mss us = − Ksb − ω2 Msb ub , being : ub =
1
u (3) ul + ui−l + t*lk uk dΓ = u*lk tk dΓ (5)
0 ff 2 Γ Γ

3
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107

∫ ∫
1 ( i+ ) velocity, nj is the unit normal at the boundary point, and eiωt has been
t − ti−l + s*lk uk dΓ = d*lk tk dΓ (6)
2 l Γ Γ omitted for brevity.

k , t k and uk , t k are displacements and tractions at the collo­

where ui+ i+ i− i−
cation point along respectively the positive and negative crack faces,
and d*lk and s*lk are obtained from the differentiation of u*lk and t*lk (see e.g.
Domínguez et al., 2000). In Eqs. 5,6, it has been assumed that the
collocation point xi is located at a smooth boundary point
( )
(Γshm xi ∈ C 1 ) leading to the 1/2 factor present in them. This
assumption is related to the required use of C 1 geometric continuity at
collocation points for Hypersingular BIEs. In order to overcome this, the
Multiple Collocation Approach (MCA) (Ariza and Domínguez, 2002) is
used when collocating at a crack boundary point. Standard quadratic
triangular (6 nodes) and quadrilateral (9 nodes) boundary elements are
considered for the discretization.
The shell region Ωsh is discretized by using the FEM. Shell finite el­
ements based on the degeneration from the three-dimensional solid are
considered (Ahmad et al., 1970). The shear and membrane locking
phenomena related to these elements are overcome by using the Mixed
Interpolation of Tensorial Components (MITC) proposed by Bathe and
co-workers. The MITC9 shell finite element (Bucalem and Bathe, 1993)
is used in this work. The equilibrium equation for a given shell finite
element l can be written as:
̃ (l) a(l) − Q(l) t(l) = f (l)
K (7)
̃ (l) = K(l) − ω2 M(l) is the resulting time harmonic stiffness matrix,
where K
a(l) is the vector of nodal displacements and rotations, Q(l) is the matrix
transferring distributed mid-surface load t(l) to nodal loads, and f (l) is the
vector of equilibrating nodal forces.
A conforming mesh between the crack-like boundary and the shell
mid-surface is considered. Then, the coupling is performed by imposing
perfectly welding conditions through the following compatibility and
equilibrium between shell finite element and crack-like boundary:
u+ − l
k = uk = uk
t+ − l
k + tk + tk = 0
2.3. Soil-beam model
For a further reduction in the number of degrees of freedom of the
problem, the monopile shaft or suction caisson skirt can be simplified to
a unidimensional beam element. By doing so, the soil-foundation
interface is concentrated into the beam axis and the behaviour at each
point of the foundation element is defined by the cross-section dis­
placements and rotations and the resultant of the interaction tractions
along the ring. Evidently, this kind of model is not able to capture the
local effects produced at the soil-shell interface, but they can accurately
approximate the global response of the foundation (especially for
medium-to-large aspect ratios).
A numerical model with these features was proposed by the Authors
in a previous work (Álamo et al., 2016) for the analysis of pile founda­
tions. In addition to the reduction in the number of degrees of freedom
achieved by the omission of the soil-shell interface, the developed model
makes use of Green’s functions for the layered half space instead of the
previous full-space fundamental solution. Thus, the discretization of the
soil free-surface and any strata interface is avoided as the Green’s
functions already satisfy the boundary conditions of those contours. As
result, an efficient numerical model is obtained in which the only vari­
ables correspond to the mid-line of the foundation element. In the
following, the basis of the model formulation are outlined. For a more
detailed description, the original work (Álamo et al., 2016) is referred.
The soil region Ωs is assumed to be formed by, in general, a group of
horizontal layers overlying a half space. The presence of the foundation
element is represented through a load line Γb over which the distributed
soil-foundation interaction forces act, being these the only body forces in
the soil domain. Due to the treatment of the foundation as a load line,
the only boundary of the soil domain correspond to the free-surface Γfs .
Considering that its zero-traction boundary condition is already satisfied
by both the Green’s functions and the unknown state, the Singular BIE
for internal points of the soil domain can be reduced to:
(8a)
∫
uil = u*lk qk dΓb
̃ (11)
(8b) Γb

where ulk denotes the shell displacements and tlk the distributed mid- where ̃
*
ulk is the displacement Green’s function for the layered half space
surface shell load. proposed by Pak and Guzina (Pak and Guzina, 2002) and qk are the
The seismic input is included in the formulation by following the distributed interaction forces acting over the soil. In order to numeri­
classical decomposition of the total field into the superposition of the cally evaluate the line integral, classic lineal (2 nodes) elements are
incident field (produced by the impinging seismic excitation) and the considered. Also, a particular non-nodal collocation strategy over four
scattered field (produced by the foundation), see e.g. (Domínguez, points of the fictitious soil-shell interface is required in order to avoid
1993): the singularity of the Green’s function (see Álamo et al., 2016 for more
utot in sc
(9a) details).
k = uk + uk
The finite element modelling of the foundation piece is done by using
ttot in sc
(9b) two-noded beam elements. Cubic and quadratic shape functions that
k = t k + tk
satisfy the Timoshenko’s beam static equation (Friedman and Kosmatka,
Since the displacements and tractions present in the BIEs have to be 1993) are used for the lateral behaviour, while linear shape functions are
evanescent, they are substituted by the scattered field, which, under the used to model the distributed interaction forces and axial displacements.
previous assumption, is equivalent to subtracting the incident field from The equilibrium equation for a given beam element l can be written as:
the total field. The considered incident field is a simple vertically inci­
dent shear wave, which has the following non-zero displacements and
̃ (l) a(l) − Q(l) q(l) = f (l)
K (12)
tractions:
Note that the terms of this equation are the beam-counterparts of the
1( ) ones presented in Eq. (7), being q(l) the vector defining the nodal values
uin = e− iks x3
+e iks x3
(10a)
1
2 of the distributed interaction forces acting over the foundation element.
Conforming meshes are considered to discretize the soil load line and
tin in
1 = Gs u1,3 n3 (10b) the foundation element. Thus, the coupling between both regions can be
easily done by, again, imposing compatibility and equilibrium condi­
tin in
3 = Gs u1,3 n1 (10c) tions in terms of displacements and interaction forces, respectively:
√̅̅̅̅̅̅̅̅̅̅̅̅ uk = ulk (13a)
where ks = ω/cs is the S wavenumber, cs = G*s /ρs is the shear wave

4
G.M. Álamo et al.
Fig. 3. Comparison between static stiffnesses.
qk + qlk = 0 (13b)
where ulk denotes the beam displacements and qlk the distributed inter­
action forces acting over the beam.
Finally, and following the same strategy than in the previous soil-
shell model, the seismic excitation is introduced by superposing the
incident and scattered fields. Note that, owing to the reduced expression
of the Singular BIE, in the soil-beam model only the displacement terms
of the incident field are necessary.
3. Results and discussion
In this section, a comprehensive comparative study between results
from both models is given. It includes the necessary ingredients for
comparing the foundation characterization (impedances and kinematic
interaction factors) as well as the final OWT variables of interest
(bending moments, shear forces, displacements and rotations).
The study covers the following parameters: length to diameter ratio:
L/D = {1, …, 10}, thickness to diameter ratio: t/D =
{0.001, 0.01, 0.02}, and foundation shear modulus to soil shear
modulus ratio: Gf /Gs = {1000, 4000, 16000}.
For this class of foundations in homogeneous soils this set of pa­
rameters completely defines the problem. Nonetheless, Doherty et al.
(2005) found that the relative stiffness between the foundation lateral
5
Computers and Geotechnics 134 (2021) 104107
shell (bucket skirt or pile shaft) and the soil can effectively be synthe­
sized via a dimensionless parameter JDoherty = (Ef t)/(Gs R) which relates
shell membrane stiffness and soil stiffness (Ef is the Young’s modulus of
the foundation and R is the radius of the foundation cross-section). It
allows an approximate but useful reduction of the number of defining
parameters. For convenience’s sake, this dimensionless parameter is
defined as J = (Gf t)/(Gs D) in the present paper, which differs from the
original by the factor JDoherty /J = 4(1 +νf ) (constant since νf = 0.25 in all
cases). Although the work of Doherty et al. (2005) is limited to static
stiffnesses, it is shown in Appendix A that, in general, J remains as a
useful parameter in dynamics for frequencies ao = fD/cs < 0.3 and
t/D ≤ 0.05.
The rest of dimensionless parameters that define the studied prob­
lems are: hysteretic damping ratios of foundation ξf = 2% and soil ξs =
5% materials, foundation soil density ratio ρf /ρs = 3.9, and soil Pois­
son’s ratio νs = 0.49 (saturated soil).
3.1. Static stiffnesses
Static stiffnesses are relevant for the calculation of the fundamental
frequency of OWTs since it is usually small, typically f1 ∼ 0.3 Hz (Kay­
nia, 2018). For seismic analyses, however, the use of static stiffnesses is
generally not recommended.
Fig. 3 shows the vertical, horizontal, rocking and sway-rocking cross-
G.M. Álamo et al.
Fig. 4. Relative errors for static stiffnesses (νs = 0.49).
coupling stiffness components for most of the cases studied. One case per
t/D has been removed since they have similar values of J (and similar
graphs) to other cases. This way, the stiffness components are distrib­
uted along columns, and the relative stiffnesses between foundation and
soil are distributed along rows. Each graph shows L/D in abscissas, and
contains the results from both models. In addition to the aforementioned
Poisson’s ratio νs = 0.49 (saturated soil), a value of νs = 0.3 has been
also considered in this section because the soil is expected to behave
more similar to the drained solid in the static regime. However, the
influence of the Poisson’s ratio in the comparison between the two
models is negligible. The beam model leads to similar results to the shell
model except for very small values of J. This phenomenon is reasonable
since the validity of the beam–soil continua is kept as long as the
foundation behaves as a structural member, which happens when there
exists stiffness contrast between this element and the soil. On the other
hand, the applicability of the beam model regarding the length to
diameter ratio (L/D) is surprisingly good even at L/D = 1.
In order to give a measure of the fidelity of the soil-beam model, the
relative error between both models is given in Fig. 4. Each graph has
now the parameter J in abscissas, so that all different cases of t/D and
Gf /Gs are represented. Each length to diameter ratio is represented using
a different color: L/D = 2 (black), L/D = 4 (red), L/D = 6 (blue), L/D =
8 (green) and L/D = 10 (orange); and each thickness to diameter ratio is
shown with a different point type: t/D = 0.001 (plus), t/D = 0.01 (cross)
and t/D = 0.02 (square). In all cases, it is roughly observed that error
Fig. 5. Comparison between dynamic stiffnesses: L/D = 6, t/D = 0.01 and Gf /Gs = 4000 (J = 40).
6
Computers and Geotechnics 134 (2021) 104107
decreases as J increases, except for rocking and sway-rocking cross-
coupling stiffnesses for L/D = 2. The vertical stiffness shows a gradual
reduction of the error as L/D increases, whereas the lateral mode stiff­
nesses show little differences in the behaviour for L/D ≥ 4. Overall, the
beam model achieves errors below 5% for the lateral mode stiffnesses
and errors below 10% for the vertical stiffness when J ≥ 10 and L/D ≥ 4.
3.2. Impedance functions
The use of dynamic stiffnesses (also known as impedance functions)
allows a much more general linear SSI analysis. They can be used for the
calculation of OWT natural frequencies, as well as seismic analyses via
the substructuring procedure. In this section, the comparison between
beam and shell models is extended to impedance functions. The
frequency-dependent differences between the two models are quantified
through the relative error defined in Eq. (14). The proposed expression
compares the absolute value of the complex difference between the
result X obtained by the shell or beam approaches with respect to the
maximum value obtained by the reference model (i.e., soil-shell
approach) in the studied frequency range. This definition is preferred
over a frequency-by-frequency relative comparison in order to avoid
peak values of the error around the frequencies in which the reference
result approaches to zero.
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107

Fig. 6. Limiting frequencies alim

o,5% (error <5%) and ao,10% (error <10%) of dynamic stiffnesses.
lim

to fail beyond a quarter-wavelength per diameter (ao ≥ 0.25). In order

to give a more concise measure of this limiting frequency, two limiting
frequencies alim
o,5% and ao,10% are defined when the error reaches respec­
lim

tively 5 and 10 per cent.

Fig. 6 shows the values of alimo,5% and ao,10% (calculated as in the pre­
lim

vious illustrative case) for all cases under study. The line and point styles
are also similar to the preceding section. The previously mentioned limit
of ao ≥ 0.25 (associated with a quarter-wavelength per foundation
diameter) is a valid indicative value for obtaining errors below 5%,
although only when J ≥ 10. In the vertical and rocking components,
o,5% are located below 0.2. This erratic behaviour is nonetheless
some alim
due to the presence of a peak in the beginning of these error curves
which lightly exceed the 5% (see supplementary data). This is further
demonstrated by observing that alim o,10% do not show this behaviour.

Fig. 7. Comparison between kinematic interaction factors: L/D = 6, t/D = 0.01

and Gf /Gs = 4000 (J = 40).

|X beam (ao ) − X shell (ao )|

Δ(ao ) = [ ] (14)
max X shell (ao )
a0 ∈[0,0.5]

Fig. 5 shows the impedance curves (real and imaginary parts) as well
as the relative error for the illustrative case with L/D = 6, t/D = 0.01
and Gf /Gs = 4000, i.e. J = 40. All other cases are included as supple­
mentary data. For all stiffness components, the error between the beam
and the shell models is approximately constant initially, and then it
starts to increase with the frequency. Such effect is physically justifiable
in terms of the comparison between soil wavelength and foundation
diameter. For frequencies leading to wavelengths comparable to the
foundation diameter, the reduction to a load line performed by the beam
model becomes inadequate. Thus, it is reasonable that this model starts Fig. 8. Limiting frequencies alim
o,5% (error <5%) and ao,10% (error <10%) of ki­
lim

nematic interaction factors.

7
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107

Table 1 o,10% ∼ 0.15 ÷ 0.5. There is no clearly defined trend for the limiting
alim
Coefficients for the expressions of the limiting frequencies. Error <5%. frequencies due to the presence of the previously mentioned peaks.
c0 c1 c2 Not applicable for Nonetheless, it is reasonable to use the conventional quarter-wavelength
limiting frequency (alim
o = 0.25) for the integral model regarding kine­
KV 0.2584 − 0.0332 0.0054 J < 4, L/D < 4
KH 0.0322 0.0881 − 0.0111 J<4
matic interaction factors, which achieves errors up to approximately
KR 1.2758 − 0.4636 0.0455 J < 16, L/D < 4, Gf /Gs < 4000
10% for L/D > 2 and J > 10.
Iu 0.1135 − 0.0282 0.0065 L/D < 6 (use alim
o ≈ 0.084)
3.4. Expressions for the limiting frequencies
Iθ 0.0343 0.0157 0 L/D < 6 (use alim
o ≈ 0.066)

From the results presented in the previous sections, ready-to-use

3.3. Kinematic interaction factors
In this section, the discrepancies between both models are studied for
the kinematic interaction factors Iu and Iθ (foundation with unrestrained
head). These represent the filtering produced by the foundation in terms
of displacements and rotations at the head of the foundation (z = 0)
when subjected to an incident wave field. In this case, the study is
limited to vertically incident shear waves, which are typically the most
relevant ones.
Fig. 7 shows the kinematic interaction factors Iu and Iθ (real and
imaginary parts) as well as the relative error (defined as in the previous
section) for the illustrative case with L/D = 6, t/D = 0.01 and Gf /Gs =
4000, i.e. J = 40. All other cases are included as supplementary data. A
good agreement between both models is observed, although higher
discrepancies in the form of horizontal translation appear as the fre­
quency increases. Thus, the error curves not only increase with the
frequency, but also show several peaks and valleys. As in the case of
impedance functions, it is possible to define two limiting frequencies
alim
o,5% and ao,10% when the error reaches respectively 5 and 10 per cent.
lim
Fig. 8 shows the values of alim
o,5% and ao,10% for all cases under study.
lim
formulas are obtained in order to estimate the limiting frequencies
that ensure a certain error when using the beam model. A quadratic
polynomial in terms of the logarithm of the foundation-soil relative
stiffness parameter J is assumed for approximating the limiting
frequencies:
alim
o,e ≈ c0 + c1 logJ + c2 (logJ)
2
(15)
Tables 1 and 2 give the values of coefficients ci for the two maximum
errors of 5% and 10% respectively. Expressions are presented for all
impedance functions and kinematic interaction terms with the exception
of the sway-rocking cross-coupling impedance. This component is not
included as it is generally less restrictive than either the lateral or
rocking impedances (making no sense to use the former without the
latter). Note that linear (c2 = 0) or constant (c1 = c2 = 0) expressions
are preferred when they can be used without a significant loss in accu­
racy with respect to the quadratic formula.
The expressions provided by Tables 1 and 2 are not recommended for
values of J outside the studied interval [1, 400]. Also, for some compo­
nents, the proposed formulas are not adequate for certain limit scenarios
(such as low foundation-soil stiffness contrast), which are indicated in

The limiting frequencies are roughly alim the last column of each table. The proposed formulas (lines) together
o,5% ∼ ao,10% ∼ 0.1 for L/D = 2,
lim
with the previous results (points) are shown in Fig. 9. In each graphical
whereas for L/D > 2 and J > 10 these are alim
o,5% ∼ 0.1 ÷ 0.2 and area, the limiting frequency is plotted against the dimensionless

Table 2
Coefficients for the expressions of the limiting frequencies. Error <10%.
c0 c1 c2 Not applicable for

KV 0.2664 0.0041 0 -
KH 0.2760 0 0 J<4
KR 0.9667 − 0.2693 0.0262 J < 10, Gf /Gs < 4000
Iu 0.5248 − 0.0462 0.0026 L/D < 6 (use alim
o ≈ 0.13 [L/D = 2] or 0.32 [L/D = 4])
Iθ 0.0414 0.0335 0 L/D < 4 (use alim
o ≈ 0.10)

Fig. 9. Comparison between the obtained limiting frequencies (points) and their proposed expressions (lines).

8
G.M. Álamo et al.
Fig. 10. Frequency Response Functions for representative variables of the OWT system. Comparison between the different models.
parameter J. Each column corresponds to a different foundation vari­
able, while each row corresponds to a different maximum permitted
error. Note that the points outside the applicability range of the obtained
expressions are not included in the figure. For the kinematic interaction
factors, the alternative values for the points outside the applicability
range are also shown in red and purple colours.
3.5. Application example
In order to illustrate the accuracy of the soil-beam model to repro­
duce the response of the OWT-foundation system, a practical example is
briefly presented in this section. The system is based on the reference
NREL-5 MW OWT model (Jonkman et al., 2009). The tower is 70 m high
and it has a hollow cross-section with variable diameter from 6 m at its
base to 3.87 m at the hub height, whereas a constant thickness to
diameter ratio of 0.45% is assumed. The supporting structure is a 20 m
high, 6 m diameter and thickness ratio of 1% tubular member. No
transition piece between supporting structure and tower is considered in
the analyses. The foundation presents the same diameter (D = 6 m) and
9
Computers and Geotechnics 134 (2021) 104107
thickness ratio (t/D = 1%) as the supporting structure. An intermediate
embedment length ratio L/D = 6 is assumed in this example. The whole
structure (tower, supporting monopile and foundation) is considered to
be made of steel with: Young’s modulus Ef = 210 GPa, density ρf = 7850
kg/m3 and Poisson’s ratio νf = 0.25. For the superstructure a hysteretic
damping coefficient ξf = 2.5% is considered.
The soil properties are selected in order to reproduce a saturated
media through elastic equivalent properties: density ρs = 2000 kg/m3 ,
Poisson’s ratio νs = 0.49 and hysteretic damping coefficient ξs = 5%.
Different foundation-soil stiffness ratios J = 10, 40 and 160 are assumed
in order to cover the range studied in the previous sections. These values
correspond to soils whose shear wave velocities are approximately cs =
200, 100 and 50 m/s, respectively.
In coherence with the obtained kinematic interaction factors, the
considered seismic excitation is a vertically incident shear wave. The
free-field displacement is denoted as uff , while its acceleration is üff .
Fig. 10 shows the Frequency Response Functions (FRF) of several
representative variables of the OWT system. The second and third rows
present the FRF of the displacement and rotation atop the turbine tower
G.M. Álamo et al.
Fig. 11. Validity of J for characterizing the relative foundation-soil stiffness in the computation of impedance functions.
with respect to the free-field displacement. On the other hand, the fourth
and fifth rows illustrate the shear force and bending moment at the base
of the supporting structure (mud-line level) with respect to the free-field
acceleration. The results obtained by assuming a fixed-base model (no
SSI) are compared with the ones obtained through the substructuring
procedure (see Section 2.1) using both the soil-shell or soil-beam models
to characterize the foundation. In order to help in interpreting the re­
sults, the first row shows the limiting frequencies (corresponding to a
maximum relative error between the shell and beam models less than
10%) for all of the foundation variables involved in the problem at hand.
Each column of the figure corresponds to a different soil-foundation
relative stiffness ratio. Frequencies up to 20 Hz are considered as a
wide frequency range for the energy content of the seismic excitation.
The results presented in Fig. 10 show a clear influence of the SSI
effects on all studied variables. In general terms, introducing the foun­
dation behaviour results in a higher response of the system for small
frequencies (< 5 Hz), and a lower response for higher frequencies. The
shifting of the system natural frequencies toward lower values produced
by the SSI effects is also seen in the obtained results.
The accuracy of using the beam model instead of the shell one can be
also tested by comparing their results in Fig. 10. A good agreement
between both models is found for all variables. Some discrepancies are
produced for frequencies above certain limiting frequencies. The hori­
zontal impedance term seems to be the most important one, followed by
the limit corresponding to the lateral kinematic interaction factor.
However, these differences between the soil-shell and soil-beam models
10
Computers and Geotechnics 134 (2021) 104107
are not significant for the example case studied (note the logarithmic
scale in the forces at the base variables). Thus, the results show that the
real factor that limits the use of the soil-beam model is the maximum
value of the dimensionless frequency ao = 0.5, i.e., foundation diameter
equal to half of the wavelength. This restriction makes the soil-beam
model to be used with caution for seismic analyses of large diameter
foundations in extremely soft soils. The validity of the beam model in
this scenario will strongly depend on the frequency content of the
excitation, and the evolution with frequency of the studied FRF.
4. Conclusions
This paper presents a comparison between soil-shell and soil-beam
models for the dynamic characterization of OWT foundation elements
(piles or suction caissons) and the seismic analysis of the complete
system. The reference results are computed with the first model, that
uses Boundary Elements to model the soil behaviour coupled with shell
Finite Elements to represent the foundation element. On the other hand,
the soil-beam model is based on the integral expression of the Reci­
procity Theorem together with advanced Green’s functions for the soil
modelling, while the structural behaviour of the foundation is handled
via beam Finite Elements. For the analyses, soil and foundation
geometrical and material properties typical of this singular construction
are assumed.
First, the comparison is made in terms of the foundation character­
ization variables: static stiffness, impedance functions and kinematic
G.M. Álamo et al.
Fig. 12. Validity of J for characterizing the relative foundation-soil stiffness in the computation of kinematic interaction factors.
interaction factors. A dimensionless analysis is made in order to present
more general results. The main conclusions drawn from this study are:
• The dimensionless parameter J proposed by Doherty et al. (2005) to
define the relative stiffness between the foundation and soil in static
can be also applied for dynamic analyses.
• In general terms, the soil-beam model accurately reproduces the
global foundation response with respect to the rigorous soil-shell
model. The agreement is quite good even for foundations with
small aspect ratios.
• The accuracy of the beam model is reduced for high frequencies and
low foundation-soil stiffness contrast.
• Closed-form expressions are proposed in order to estimate the
applicability range, in terms of maximum dimensionless frequency,
of the beam simplification. Those are functions of the foundation-soil
relative stiffness parameter J and depend on the foundation variable
to compute and maximum admissible error.
An application example is also presented, in which the seismic
response of a 5 MW OWT including SSI effects is computed via a sub­
structuring procedure. The FRF of key structural variables obtained by
both models (shell and beam) are compared, and the main conclusions
drawn from this example are the following:
• The soil-structure interaction effects significantly change the seismic
response of the OWT system.
• The foundation variable whose modelling has more impact on the
obtained results is the lateral impedance term, followed by the
lateral kinematic interaction factor.
• Below the proposed limiting frequencies, virtually the same results
are obtained regardless using the beam or shell model.
11
Computers and Geotechnics 134 (2021) 104107
• Even above these limits, the foundation beam model accurately re­
produces the OWT response in an acceptable frequency range for
seismic analyses. The upper bound corresponds to the frequency for
which the foundation diameter coincides with half wavelength.
• This restrains the use of the beam model for extremely soft soils if the
high frequency content of the excitation is important. But for typical
scenarios, the foundation beam simplification is a valid option for
reproducing the OWT seismic response, making it a valuable tool
especially for design or optimization steps.
CRediT authorship contribution statement
Guillermo M. Álamo: Conceptualization, Methodology, Software,
Investigation, Writing - original draft. Jacob D.R. Bordón: Methodol­
ogy, Software, Investigation, Writing - original draft, Writing - review &
editing. Juan J. Aznárez: Supervision, Funding acquisition.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Acknowledgements
This work was supported by Agencia Estatal de Investigación (AEI) of
Spain through research project BIA2017-88770-R. J.D.R. Bordón is a
postdoctoral fellow of the ULPGC Postdoctoral Programme. The authors
are grateful for this support.
G.M. Álamo et al. Computers and Geotechnics 134 (2021) 104107

Appendix A. Validity of J as characteristic dimensionless parameter for dynamic analyses

The dimensionless parameter J was proposed by Doherty and Deeks (2003) to characterize the relative stiffness between the foundation and soil for
static analyses. In this appendix, its use in dynamic regime is tested by computing the impedance functions and kinematic interaction factors for
several configurations and comparing the results obtained for the same value of J.
Five shell thickness ratios t/D = {0.001, 0.01, 0.02, 0.05, 0.1} are combined with a continuum range of values for the ratio between the foundation
and soil shear modulus Gf /Gs in order to cover a comparable J interval. For brevity’s sake, only results for a configuration with aspect ratio L/D = 6 are
presented. The rest of dimensionless properties are equal to the ones defined in Section 3.
Fig. 11 plots the impedance functions against the parameter J for the studied configurations. Real and imaginary components are presented in pairs
for several dimensionless frequencies distributed in rows. Each column correspond to different impedance modes. The results show that the parameter
J can be used to represent the relative foundation-soil stiffness for the static and low frequency scenarios. For higher frequencies the results are sensible
to the thickness ratio if its value if larger than 2%, especially the vertical and lateral modes. However, the foundation elements for OWT structures
typically present thickness ratios below this value, so the parameter J can be safely used to represents the foundation-soil stiffness contrast.
Fig. 12 shows now the results in terms of kinematic interaction factors. The real and imaginary components of the lateral and rotational terms are
presented in pairs of rows. Each column correspond to different dimensionless frequencies (static values are omitted as their results are trivial). As
commented before, for thickness ratios below 5%, the behaviour of the different configuration is determined by the J parameter. Thus, it can be also
used to characterize the foundation-soil relative stiffness when studying the foundation kinematic response.

Appendix B. Supplementary material

Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.compgeo.2021.104107.

References Domínguez, J., 1993. Boundary Elements in Dynamics. International Series on

Computational Engineering. Computational Mechanics Publications.
Domínguez, J., Ariza, M.P., Gallego, R., 2000. Flux and traction boundary elements
Ahmad, S., Irons, B.M., Zienkiewicz, O.C., 1970. Analysis of thick and thin shell
without hypersingular or strongly singular integrals. Int. J. Numer. Meth. Eng. 48,
structures by curved finite elements. Int. J. Numer. Meth. Eng. 2 (3), 419–451.
111–135.
Álamo, G.M., Martínez-Castro, A.E., Padrón, L.A., Aznárez, J.J., Gallego, R., Maeso, O.,
Friedman, Z., Kosmatka, J.B., 1993. An improved two-node Timoshenko beam finite
2016. Efficient numerical model for the computation of impedance functions of
element. Comput. Struct. 47 (3), 473–481.
inclined pile groups in layered soils. Eng. Struct. 126 (1), 379–390.
Gazetas, G., Dobry, R., 1984. Horizontal response of piles in layered soils. J. Geotech.
Almeida, V.S., de Paiva, J.B., 2004. A mixed BEM-FEM formulation for layered soil-
Eng. 110 (1), 20–40.
superstructure interaction. Eng. Anal. Boundary Elem. 28 (9).
Houlsby, G.T., Byrne, B.W., 2005a. Design procedures for installation of suction caissons
Arany, L., Bhattacharya, S., Macdonald, J., Hogan, S.J., 2017. Design of monopiles for
in clay and other materials. Proc. Inst. Civ. Eng. 158 (GE2), 75–82.
offshore wind turbines in 10 steps. Soil Dyn. Earthq. Eng. 92, 126–152.
Houlsby, G.T., Byrne, B.W., 2005b. Design procedures for installation of suction caissons
Ariza, P., Domínguez, J., 2002. General BE approach for three-dimensional dynamic
in sand. Proc. Inst. Civ. Eng. 158 (GE3), 135–144.
fracture analysis. Eng. Anal. Boundary Elem. 26, 639–651.
Jia, N., Zhang, P., Liu, Y., Ding, H., 2018. Bearing capacity of composite bucket
Ashuri, T., Zaaijer, M.B., Martins, J.R.R.A., van Bussel, G.J.W., van Kuik, G.A.M., 2014.
foundations for offshore wind turbines in silty sand. Ocean Eng. 151, 1–11.
Multidisciplinary design optimization of offshore wind turbines for minimum
Jiménez-Salas, J.A., Belzunce, J.A., 1965. Theoretical solution of stress distribution in
levelized cost of energy. Renewable Energy 68, 893–905.
piles. In: Proceedings of the VI International Conference on Soil Mechanics and
Bhattacharya, S., Nikitas, N., Garnsey, J., Alexander, N.A., Cox, J., Lombardi, D.,
Foundation Engineering.
Wood, D.M., Nash, D.F.T., 2013. Observed dynamic soil–structure interaction in
Jonkman, J., Butterfield, S., Musial, W., Scott, G., 2009. Definition of a 5-MW Reference
scale testing of offshore wind turbine foundations. Soil Dyn. Earthq. Eng. 54, 47–60.
Wind Turbine for Offshore System Development. Technical Report NREL/TP-500-
Bordón, J.D.R., Aznárez, J.J., Maeso, O., 2017. Dynamic model of open shell structures
38060, National Renewable Energy Laboratory.
buried in poroelastic soils. Comput. Mech. 60 (2), 269–288.
Kaynia, A., Kausel, E., 1982. Dynamic stiffness and seismic response of pile groups.
Bordón, J.D.R., Aznárez, J.J., Padrón, L.A., Maeso, O., Bhattacharya, S., 2019. Closed-
resreport R82–03, Massachusetts Institute of Technology.
form stiffnesses of multi-bucket foundations for OWT including group effect
Kaynia, A.M., 2018. Seismic considerations in design of offshore wind turbine. Soil Dyn.
correction factors. Marine Struct. 65, 326–342.
Earthq. Eng.
Bucalem, M.L., Bathe, K.J., 1993. Higher-order MITC general shell elements. Int. J.
Novak, M., Nogami, T., Aboul-Ella, F., 1978. Dynamic soil reactions for plane strain case.
Numer. Meth. Eng. 36, 3729–3754.
J. Eng. Mech. Division (ASCE) 104, 953–959.
Coda, H.B., Venturini, W.S., Aliabadi, M.H., 1999. A general 3D BEM/FEM coupling
Padrón, L., Aznárez, J., Maeso, O., 2007. BEM-FEM coupling model for the dynamic
applied to elastodynamic continua/frame structures interaction analysis. Int. J.
analysis of piles and pile groups. Eng. Anal. Boundary Elem. 31 (6), 473–484.
Numer. Meth. Eng. 46 (5), 695–712.
Page, A.M., Naess, V., de Vaal, J.B., Eiksund, G.R., Nygaard, T.A., 2019. Impact of
Cotter, O., 2009. The installation of suction caisson foundations for offshore renewable
foundation modelling in offshore wind turbines: comparison between simulations
energy structures. PhD thesis, University of Oxford - Magdalen College.
and field data. Marine Struct. 64, 379–400.
DNV, 2014. Offshore standard DNV-OS-J101: design of offshore wind turbine structures.
Pak, R.Y.S., Guzina, B.B., 2002. Three-dimensional Green’s functions for a multilayered
Det Norske Veritas.
half-space in displacement potentials. J. Eng. Mech. 128 (4), 449–461.
Doherty, J.P., Deeks, A.J., 2003. Elastic response of circular footings embedded in a non-
Poulos, H.G., Davis, E.H., 1968. The settlement behaviour of single axially loaded
homogeneous half-space. Géotechnique 53 (8), 703–714.
incompressible piles and piers. Géotechnique 18, 351–371.
Doherty, J.P., Houlsby, G.T., Deeks, A.J., 2005. Stiffness of flexible caisson foundations
Thurman, A., D’Appolonia, E., 1965. Computed movement of friction and end-bearing
embedded in nonhomogeneous elastic soil. J. Geotech. Geoenviron. Eng. 131 (12),
piles embedded in uniform and stratified soils. In: Proceedings of the VI International
1498–1508.
Conference on Soil Mechanics and Foundation Engineering, number 2, pp. 323–327.

12